Cloud layer directional scheduling method based on low earth orbit satellite and 5G network

Abstract

The invention discloses a cloud layer directional scheduling method based on a low earth orbit satellite and a 5G network, and belongs to the technical field of artificial influence weather. According to the method, a cloud electromagnetic fingerprint spectrum is generated through cross-domain fusion data of a low-orbit satellite constellation and a ground scatterometer array; deploying a federated learning framework based on 5G MEC edge nodes, and constructing a cloud layer behavior prediction model with a transfer learning capability; a low earth orbit satellite and 5G base station cooperative beam forming technology is adopted, and quantum encryption is combined to generate a scheduling instruction; constructing a multi-modal intervention effect evaluation matrix by using 5G inductance integration characteristics; constructing a cross-regional computing power pool through an inter-satellite laser link, and dynamically allocating resources based on a particle swarm optimization algorithm; a block chain technology is introduced to form a self-evolution scheduling closed loop. According to the method, accurate perception, prediction, intervention and evaluation of the cloud layer are realized, the precision, efficiency and safety of directional regulation and control of the cloud layer are improved, and technical support is provided for coping with extreme weather.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of weather modification, in particular to a cloud layer directional scheduling method based on low-orbit satellites and 5G networks. BACKGROUND

[0002] Traditional weather modification technologies (such as rockets, anti-aircraft guns for rain enhancement) have low precision, limited coverage, and rely on human experience, making it difficult to achieve directional regulation and dynamic intervention of cloud layers. At the same time, with the rapid development of 5G communication, low-orbit satellite constellation, artificial intelligence and other technologies, there is an urgent need for an intelligent cloud layer scheduling scheme that integrates space-ground networks to meet the needs of precise perception, real-time prediction and directional intervention of cloud evolution. The present application is based on this real demand, aiming to build a full-process, high-precision, self-evolving cloud layer directional scheduling system through deep collaboration between low-orbit satellites and 5G networks, to achieve directional regulation of cloud layer motion trajectory, thickness change, and precipitation probability, and to provide technical support for weather modification and improve the ability to cope with extreme weather.

[0003] Traditional cloud layer intervention technology mainly relies on ground launch devices (such as rockets, anti-aircraft guns, and ground combustion furnaces) to scatter catalysts (such as silver iodide and dry ice) into the cloud layer, to achieve the goal of rain enhancement and hail prevention by changing the microphysical process of the cloud layer. The core principle is to use catalysts to promote cloud droplet condensation or ice crystal growth, thereby triggering precipitation. This technology is mature, easy to operate, has certain effect in short-term intervention in local areas, and is relatively low in cost, making it suitable for application in areas with complex terrain and poor transportation.

[0004] However, the precision is low, it is difficult to achieve directional targeting of specific cloud layers, the range of catalyst scattering is difficult to control, and it is easy to cause resource waste or ineffective intervention; the coverage is limited, restricted by the range of launch devices, and only the cloud layer within tens of kilometers can be affected; it relies on human judgment, lacks dynamic monitoring of the real-time state of the cloud layer, and the selection of intervention timing and dosage is subjective, resulting in poor stability of the effect; it is difficult to cope with large-scale and complex cloud systems, and for cloud layers with high moving speed and complex structure, traditional technology is difficult to achieve continuous tracking and dynamic adjustment.

[0005] In recent years, with the development of remote sensing technology and computer simulation, some cloud layer intervention technologies based on satellite observation and numerical simulation have emerged. For example, high-orbit satellites are used to obtain macro data of cloud layers (such as cloud top temperature and cloud amount), combined with meteorological models to predict cloud evolution, and then unmanned aerial vehicles or aircrafts are used to scatter catalysts for intervention. In addition, some research attempts to combine radar monitoring with ground communication networks to improve the real-time performance of cloud state perception. Compared with traditional technology, the existing technology has improved in the range and precision of cloud layer monitoring. Satellite remote sensing can achieve large-scale cloud coverage, numerical simulation can provide theoretical basis for intervention strategy, and to some extent, reduces the dependence on human experience.

[0006] However, the satellite and the ground data fusion degree is low, the high-orbit satellite observation data timeliness is poor (the delay can reach dozens of minutes), it is difficult to meet the real-time scheduling demand; lack of intelligent decision mechanism, model prediction and actual intervention are out of touch, it is impossible to adjust the strategy in real time according to the dynamic change of cloud layer; communication and computing power support is insufficient, a large range of cloud layer scheduling needs massive data processing and low latency communication, and the existing network is difficult to meet the cross-regional coordination demand; security and traceability are lacking, the data in the intervention process is not formed in a closed loop management, it is difficult to evaluate the effect and trace the responsibility, and the instruction transmission exists the risk of tampering.

[0007] Therefore, the application constructs a full closed loop intelligent scheduling system of 'perception-prediction-intervention-evaluation-evolution', realizes the deep fusion of low-orbit satellite and 5G network and solves the above technical problems. SUMMARY

[0008] Based on the above technical problems, the application discloses a cloud layer directional scheduling method based on low-orbit satellite and 5G network, specifically comprising:

[0009] S1, a dynamic cloud layer topology perception network is constructed through a low-orbit satellite constellation, a three-dimensional holographic scan is performed on the cloud layer through a satellite-borne phased array radar, an electromagnetic scattering characteristic spectrum of the cloud layer particles is extracted, the cloud layer microwave response data collected by a ground scatterometer array is cross-domain fused with satellite data by using the ultra-wideband characteristic of a 5G air interface, and a cloud layer electromagnetic fingerprint spectrum is generated;

[0010] S2, a federal learning framework is deployed based on a 5G MEC edge node, each node trains a sub-model by using a locally stored historical cloud layer scheduling case, model parameter encryption aggregation is realized through a satellite link, a cloud layer behavior prediction model with migration learning ability is constructed, and a nonlinear evolution path of the cloud layer under different intervention intensities is output;

[0011] S3, a cooperative beam forming technology of a low-orbit satellite and a 5G base station is adopted, a coherent electromagnetic wave of a specific frequency band is transmitted to the target cloud layer, an intervention window is judged by monitoring the resonance frequency change of the cloud layer particles, and a scheduling instruction based on quantum key encryption is generated;

[0012] S4, the all-in-one characteristic of the 5G network is utilized to capture the acoustic feedback signal of the cloud layer after being excited in real time, a multi-modal intervention effect evaluation matrix is constructed in combination with the cloud layer optical thickness change observed by the satellite;

[0013] S5, a cross-regional scheduling computing power pool is constructed through a low-orbit satellite intersatellite laser link, a particle swarm optimization algorithm is adopted based on the chaotic characteristics of the cloud layer evolution, and the beam irradiation time of different satellites and the signal transmission power of the 5G base station are dynamically allocated;

[0014] S6, introduce blockchain technology to record the whole process scheduling data, use smart contract to automatically trigger the iteration update of cloud layer intervention strategy, form a self-evolution cloud layer directional scheduling closed loop based on satellite-ground heterogeneous network.

[0015] Preferably, in S1, when constructing a dynamic cloud layer topology aware network through a low earth orbit satellite constellation, a plurality of low earth orbit satellites form a distributed observation array according to a preset orbit, and the phased array radar carried by each satellite performs fan beam scanning on the target airspace in a time division multiplexing manner, and the azimuth and elevation angle of the radar beam are adjusted to realize three-dimensional airspace coverage. After pulse compression processing of the echo signals obtained by scanning, the scattering intensity of cloud particles at different frequency bands is extracted , which is the radar transmission frequency, combined with the propagation attenuation coefficient of radar waves in the cloud layer , the electromagnetic scattering characteristic spectrum of cloud particles is obtained by calculating the scattering cross section , where is the equivalent radius of the particle, is the distance from the satellite to the cloud layer, is the transmission power, , are the transmission and reception antenna gains respectively, is the radar wavelength.

[0016] Preferably, in S1, the super wideband characteristics of the 5G air interface are used, and the ground scatterometer array at the preset frequency band is used to realize multi-frequency band synchronous observation of the cloud layer through the super wideband carrier aggregation technology of 5G NR, and the cloud microwave response data obtained by the scatterometer at different positions is collected , where is the scatterometer number, is the observation time, the electromagnetic scattering characteristic spectrum obtained by the satellite is spatiotemporally aligned with the ground data , a fusion weight matrix is constructed , and the weight is dynamically adjusted by using the covariance correction term, where , are the weight coefficients of the satellite data and the ground data respectively, is the frequency point index, , are the mean values of the satellite and ground data respectively, is the variance, and the cloud electromagnetic fingerprint spectrum is generated, and the formula is: , realizing the cross-domain fusion of satellite and ground data.

[0017] Preferably, in step S2, when deploying the federated learning framework based on 5G MEC edge nodes, a federated learning client is deployed in each 5G MEC edge node within the target area, and each client calls the locally stored historical cloud scheduling case dataset. ,in For the first The node of the first Each case input feature includes the initial cloud state and intervention parameters. To correspond with the output results, a local sub-model is trained using the stochastic gradient descent algorithm. Calculate the gradient ,in For the first The loss function for each node. For single sample loss, For the first The sample size of each node is [number], and after training, each node encrypts the sub-model parameters using a symmetric encryption method based on the SM4 algorithm via a 5G network. The encryption parameters are [details]. ,in The session key, pre-distributed via satellite link, is sent to a low-Earth orbit satellite acting as an aggregation server. The satellite receives the encryption parameters from all nodes and decrypts them to obtain the desired key. The aggregation model is calculated using the following formula: ,in The total number of edge nodes. For the first The weight coefficients of the sub-model parameters of each edge node during aggregation, and the weight coefficients of the sub-model parameters during aggregation, and the weight coefficients of the sub-model parameters of each edge node ... After encryption, the data is broadcast to each 5G MEC edge node via satellite link to complete the sub-model update.

[0018] Preferably, in step S2, when constructing a cloud behavior prediction model with transfer learning capabilities, the global model obtained by federated learning aggregation is used. Based on the model, using the transfer learning module, through the feature mapping function Source domain data features Features of target domain data Mapping to shared feature space Computational domain difference loss The formula is: ,in For domain discriminator, For domain tags, , The sample sizes of the source and target domains are respectively, and their values ​​are compared with the prediction loss. The total loss is obtained by weighted summation. ,in These are the model's predicted values. These are measured values. is a weight coefficient, by minimizing optimizing model parameters, model input is different intervention intensity , wherein is the first intensity corresponding to the electromagnetic wave power and the action time, the output of the cloud layer at the future moment of the nonlinear evolution path , wherein is the position coordinate, is the thickness, is the particle concentration.

[0019] Preferably, in S3, the low-orbit satellite and the 5G base station adopt the cooperative beam forming technology, the low-orbit satellite and the 5G base station covering the target cloud layer area form a distributed transmitting array, the satellite and the base station transmitting clock deviation is controlled within a preset time range through a satellite-ground time synchronization module, and the beam forming weight vector of each transmitting node is calculated based on the three-dimensional coordinates of the target cloud layer , wherein is the total number of cooperative nodes, and the formula is: , is the distance between the first node and the reference point, is the beam pointing angle, is the electromagnetic wave wavelength), control each node to synchronously transmit coherent electromagnetic waves of different frequency bands and form energy focusing in the target cloud layer area, real-time monitor the scattering echo signals of cloud particles to different frequency electromagnetic waves, and extract the resonance peak frequency in the echo spectrum through fast Fourier transform , when stabilize in a preset frequency interval , the middle is the theoretical resonance frequency, is the allowed deviation value, and the duration exceeds the threshold value, it is determined as an effective intervention window, a scheduling instruction containing beam parameters, action time, power level is generated by the low-orbit satellite, an encryption key is generated by using quantum key distribution technology , after the instruction is encrypted by , it is issued to the cloud layer intervention execution terminal through the 5G secure channel, wherein is the AES encryption algorithm, is the plaintext instruction.

[0020] Preferably, in S4, the all-sensing integration characteristics of the 5G network are used, the all-sensing antenna array of the 5G base station is used to capture the acoustic feedback signal generated after the cloud layer is excited in real time, and the acoustic characteristic spectrum is obtained by filtering and Fourier transform on the signal , wherein is the frequency, is the time, and the peak frequency in the characteristic spectrum is extracted Signal strength and duration As an acoustic evaluation metric, optical thickness data of clouds during the intervention process were obtained using a multispectral imager mounted on a low-Earth orbit satellite. Calculate the rate of change of optical thickness ,in The initial optical thickness before intervention. Using spectral wavelengths, acoustic parameters are spatiotemporally aligned with optical thickness change rates to construct a multimodal intervention effect evaluation matrix. The matrix elements are: ,in and These are the weighting coefficients for acoustic characteristics and the rate of change of optical thickness, respectively. For the first The first time window, the first Normalized values ​​of acoustic characteristics in each frequency band The normalized value of the optical thickness change rate corresponding to the time window and wavelength band is obtained through the matrix eigenvalues. To assess the overall strength of the intervention effect, when When determining the effectiveness of the intervention, among which This is a preset threshold.

[0021] Preferably, in step S5, a cross-regional scheduling computing power pool is constructed through inter-satellite laser links between low-Earth orbit (LEO) satellites. Within the LEO satellite constellation, laser communication terminals are mounted to establish communication links with other satellites, and the link transmission rate is [not specified]. ,in As the reference rate, Atmospheric attenuation coefficient, To account for the distance between satellites, each satellite will utilize the idle computing power of its computing units. By abstracting virtualization technology into schedulable computing resource blocks, a computing pool is constructed with a total capacity of [missing information]. ,in The number of satellites participating in the network. For the first The computing power utilization coefficient of each satellite is obtained by collecting the computing power requirements of cloud scheduling tasks in various regions through the onboard software-defined network controller. Based on task priority coefficient ,according to Dynamically allocate computing resources to achieve efficient operation of cross-regional computing pool scheduling, among which... For the first The satellite to the first The weighting of each task. This represents the distance between the satellite and the mission area.

[0022] Preferably, the chaos characteristic based on cloud evolution in S5 adopts a particle swarm optimization algorithm, and the beam irradiation time of the low earth orbit satellite and the signal transmission power of the 5G base station As an optimization variable, a particle position vector is constructed , wherein is the number of satellites, is the number of base stations, and the intervention effect bias A fitness function is constructed , wherein is the target intervention effect at the moment, is the actual effect, is the intervention duration, and the particle swarm velocity and position are initialized and The particle is updated, wherein is the cloud evolution chaos factor, is the inertia weight, is the learning factor, is a random number, is the particle historical optimum, is the global optimum, is the chaotic disturbance term, and iteration is performed until , the optimal beam irradiation time and signal transmission power distribution scheme are output, is the precision threshold.

[0023] Preferably, the blockchain technology is introduced in S1 to record the whole-process scheduling data, and the cloud initial state data collected by the low earth orbit satellite and the 5G network, the federated learning model parameters, the beam forming parameters, the intervention instructions, and the cloud state feedback data are timestamped to generate encrypted data blocks, and each data block contains the hash value of the previous block , a chain structure is constructed by , wherein is the current block data, is the timestamp, and the distributed nodes composed of the satellite constellation and the 5G base station jointly participate in data verification and block accounting to ensure data tamper-proofing and traceability, and at the same time, a smart contract is deployed in the blockchain, and the preset contract triggering condition is that the eigenvalue of the multi-modal intervention effect evaluation matrix is less than the threshold value for consecutive multiple monitoring periods , the particle swarm optimization algorithm is automatically called to regenerate the intervention strategy parameters, and is transmitted to the execution terminal through a quantum encryption link, and the whole-process data after the execution of the new strategy is written into the blockchain to form a new block, and through the cycle of data chaining-contract monitoring-strategy updating-data re-chaining, a self-evolution cloud directional scheduling closed loop based on satellite-ground heterogeneous network is formed.

[0024] Compared with the prior art, the technical scheme of the application has the following technical effects:

[0025] The application generates a high-precision cloud layer electromagnetic fingerprint map through the cooperative perception of low-orbit satellites and 5G networks, realizes accurate characterization of the physical characteristics of the cloud layer, comprehensively captures details such as the particle distribution and motion state of the cloud layer, provides reliable basis for subsequent directional intervention, greatly improves the accuracy and real-time performance of cloud layer perception, and solves the problem of fuzzy cloud state grasping by traditional methods.

[0026] The cloud layer behavior prediction model constructed by federated learning and transfer learning can accurately output the cloud layer evolution path under different intervention intensities, effectively improves the prediction ability of the complex cloud layer system by using distributed training and cross-scene knowledge transfer, overcomes the defect of poor generalization of traditional models, provides scientific guidance for intervention strategy formulation, and improves the forward-looking and pertinence of cloud layer regulation.

[0027] The cooperative beam forming technology of low-orbit satellites and 5G base stations in the application, combined with quantum encryption scheduling instructions, realizes directional and accurate intervention on the target cloud layer, dynamically adjusts the beam parameters and encrypted transmission instructions, ensures that the electromagnetic wave energy is focused on the target area, and ensures the security of the instruction transmission, solves the problems of low precision and insufficient safety of traditional intervention methods, and improves the effectiveness and reliability of cloud layer intervention.

[0028] The self-evolution closed loop constructed by the blockchain and the smart contract realizes the traceability of scheduling data and the automatic iteration of strategies, the multi-modal evaluation matrix quantifies the intervention effect in real time, and the resource optimization of the cross-region computing power pool makes the system can dynamically adjust the strategy, forms a closed loop mechanism for continuous optimization, overcomes the limitations of the lag of traditional technology effect evaluation and the passive strategy update, realizes the intelligentization and self-automation of cloud layer scheduling.

[0029] The above description is only a summary of the technical scheme of the application, in order to more clearly understand the technical means of the application, so as to be implemented according to the content of the specification, and in order to make the above and other purposes, characteristics and advantages of the application more obvious and easy to understand, the following will be described in detail with the preferred embodiments of the application and the accompanying drawings.

[0030] According to the detailed description of the specific embodiments of the application in the following text combined with the drawings, those skilled in the art will more clearly understand the above and other purposes, advantages and characteristics of the application. BRIEF DESCRIPTION OF DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings described below are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual proportion.

[0032] According to the description of the drawings in the file and the corresponding technical content, the titles of the drawings are as follows

[0033] Figure 1 For cloud layer directional scheduling method based on low earth orbit satellite and 5G network;

[0034] Figure 2 For cumulonimbus directional intervention system architecture based on low earth orbit satellite and 5G base station cooperative beam forming;

[0035] Figure 3 For cumulonimbus directional intervention resonant frequency monitoring curve based on low earth orbit satellite and 5G base station cooperative beam forming;

[0036] Figure 4 For stratocumulus directional intervention system architecture based on low earth orbit satellite and 5G base station cooperative beam forming;

[0037] Figure 5 For stratocumulus directional intervention resonant frequency monitoring curve based on low earth orbit satellite and 5G base station cooperative beam forming;

[0038] Figure 6 For cross-regional cloud layer scheduling computing power pool architecture based on low earth orbit satellite inter-satellite laser link;

[0039] Figure 7 For particle swarm optimization algorithm iteration optimization curve in cross-regional cloud layer scheduling. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. In the following description, specific details such as specific configurations and components are provided only to help understand the embodiments of the present application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. In addition, in order to be clear and concise, the description of known functions and structures is omitted in the embodiments.

[0041] It should be understood that the reference herein to "one embodiment" or "the embodiment" means that a particular feature, structure, or characteristic described is included in at least one embodiment of the application. Therefore, appearances of the phrases "one embodiment" or "in one embodiment" at various places in the specification are not necessarily referring to the same embodiment. Furthermore, the particular features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0042] In addition, reference numerals and / or letters can be repeated in different instances throughout the specification. Such repetition is for the purpose of simplicity and clarity and does not in itself dictate a relationship between the various embodiments and / or configurations discussed.

[0043] The term "and / or", merely describes an associated relationship between associated objects, which means that there can be three relationships, for example, A and / or B, which means that there are three cases of A alone, B alone, and A and B together. The term "and" herein is to describe another associated relationship between associated objects, which means that there can be two relationships, for example, A and B, which means that there are two cases of A alone and A and B together. In addition, the character " / " herein generally represents an "or" relationship between the associated objects before and after it.

[0044] The term "at least one" herein merely describes an associated relationship between associated objects, which means that there can be three relationships, for example, at least one of A and B, which means that there are three cases of A alone, A and B together, and B alone.

[0045] It should also be noted that the relationship terms such as first and second, etc. used herein are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion.

[0046] Embodiment 1

[0047] This embodiment mainly describes a cloud layer directional scheduling method based on low-orbit satellite and 5G network, as shown in Figure 1 Specifically, it includes:

[0048] S1, constructing a dynamic cloud layer topology perception network through a constellation of low-orbit satellites, performing three-dimensional holographic scanning on the cloud layer through a satellite-borne phased array radar, extracting electromagnetic scattering characteristic spectrum of cloud layer particles, using the ultra-wideband characteristics of 5G air interface, cross-domain fusion of cloud layer microwave response data collected by a ground scatterometer array and satellite data, and generating a cloud layer electromagnetic fingerprint map;

[0049] S2, based on the 5G MEC edge node deployment of federated learning framework, each node utilizes the local storage of historical cloud layer scheduling case training sub-model, through satellite link realizes model parameter encryption aggregation, constructs the cloud layer behavior prediction model with migration learning ability, outputs the nonlinear evolution path of cloud layer under different intervention intensity;

[0050] S3, using the cooperative beam forming technology of low earth orbit satellite and 5G base station, transmitting specific frequency band coherent electromagnetic wave to the target cloud layer, judging the intervention window by monitoring the resonance frequency change of cloud particle, generating scheduling instruction based on quantum key encryption;

[0051] S4, using the sensing integration characteristics of 5G network, real-time capturing the acoustic feedback signal of cloud layer after excitation, combining with the optical thickness change of cloud layer observed by satellite, constructing multi-modal intervention effect evaluation matrix;

[0052] S5, through the inter-satellite laser link of low earth orbit satellite, constructing cross-regional scheduling computing power pool, based on the chaotic characteristics of cloud evolution, using particle swarm optimization algorithm, dynamically allocating the beam irradiation time of different satellites and the signal transmission power of 5G base station;

[0053] S6, introducing blockchain technology to record the whole process of scheduling data, using smart contract to automatically trigger the iterative update of cloud layer intervention strategy, forming a self-evolution cloud layer directional scheduling closed loop based on satellite-ground heterogeneous network.

[0054] Further, in S1, when constructing dynamic cloud layer topology perception network through low earth orbit satellite constellation, multiple low earth orbit satellites form a distributed observation array according to the preset orbit, the phased array radar carried by each satellite scans the target airspace in the form of fan beam through time division multiplexing, by adjusting the azimuth and pitch angle of radar beam, three-dimensional airspace coverage is realized, after pulse compression processing of the echo signal obtained by scanning, the scattering intensity of cloud particle at different frequency bands , is extracted, combined with the propagation attenuation coefficient of radar wave in cloud layer , the electromagnetic scattering characteristic spectrum of cloud particle is obtained by calculating the scattering cross section , where is the equivalent radius of particle, is the distance from satellite to cloud layer, is the transmission power, , are the transmitting and receiving antenna gains respectively, is the radar wavelength.

[0055] Further, in S1, the ultra-wideband characteristics of the 5G air interface are utilized to realize multi-band synchronous observation of the cloud layer through the ultra-wideband carrier aggregation technology of the 5G NR on the ground scatterometer array in the preset frequency band, and the microwave response data of the cloud layer obtained by the scatterometers at different positions is collected wherein is the scatterometer number, is the observation time, and the electromagnetic scattering feature spectrum obtained by the satellite is spatiotemporally aligned with the ground data , a fusion weight matrix is constructed , the covariance correction term is used to dynamically adjust the weight, wherein , are the weight coefficients of the satellite data and the ground data, respectively, is the frequency point index, , are the mean values of the satellite data and the ground data, respectively, is the variance, a cloud electromagnetic fingerprint spectrum is generated, and the formula is: , realizing the cross-domain fusion of satellite and ground data.

[0056] Further, in S2, when deploying a federated learning framework based on the 5G MEC edge node, a federated learning client is deployed in each 5G MEC edge node in the target area, each client calls the historical cloud layer scheduling case data set stored locally wherein is the th node th case input feature, including the initial state of the cloud layer and the intervention parameter, is the corresponding output result, a local sub-model is trained using the stochastic gradient descent algorithm , and the gradient is calculated wherein is the loss function of the th node, is the loss of a single sample, is the sample size of the th node, after the training is completed, each node encrypts the sub-model parameters using a symmetric encryption method based on the SM4 algorithm, and the encryption parameters are wherein is the session key pre-distributed through the satellite link, which is sent to the low-orbit satellite as an aggregation server, and the satellite decrypts the encrypted parameters to obtain , and the aggregation model is calculated, and the formula is: wherein is the total number of edge nodes, is the The weight coefficients of the sub-model parameters of each edge node during aggregation, and the weight coefficients of the sub-model parameters during aggregation, and the weight coefficients of the sub-model parameters of each edge node ... After encryption, the data is broadcast to each 5G MEC edge node via satellite link to complete the sub-model update.

[0057] Furthermore, when constructing a cloud behavior prediction model with transfer learning capabilities in S2, the global model obtained through federated learning is used. Based on the model, using the transfer learning module, through the feature mapping function Source domain data features Features of target domain data Mapping to shared feature space Computational domain difference loss The formula is: ,in For domain discriminator, For domain tags, , The sample sizes of the source and target domains are respectively, and their values ​​are compared with the prediction loss. The total loss is obtained by weighted summation. ,in These are the model's predicted values. These are measured values. For the weighting coefficients, minimize Optimize model parameters; model input consists of different intervention intensities. ,in For the first The electromagnetic wave power and duration corresponding to the intensity will affect the output cloud layer in the future. Nonlinear evolution path at time step ,in For position coordinates, For thickness, This represents the particle concentration.

[0058] Furthermore, when using the collaborative beamforming technology of low-Earth orbit satellites and 5G base stations in S3, the low-Earth orbit satellites and the 5G base stations covering the target cloud area form a distributed transmission array. The satellite-to-ground time synchronization module controls the transmission clock deviation between the satellite and the base station within a preset time range, and calculates the beamforming weight vector of each transmission node based on the three-dimensional coordinates of the target cloud. ,in The total number of collaborating nodes is given by the formula: , For the first The distance between each node and the reference point Beam pointing angle, The coherent electromagnetic waves of different frequency bands are synchronously emitted by each node under the control of the system, and energy focusing is formed in the target cloud layer area. The scattering echo signals of the cloud particles to the electromagnetic waves of different frequencies are monitored in real time. The resonance peak frequency in the echo spectrum is extracted by fast Fourier transform When The preset frequency interval is stabilized , the middle The theoretical resonance frequency is The allowable deviation value is determined as an effective intervention window when the duration exceeds the threshold value. The scheduling instruction containing the beam parameter, action duration, power level is generated by the low-orbit satellite, and the encryption key is generated by using the quantum key distribution technology After the instruction is encrypted, it is issued to the cloud layer intervention execution terminal through the 5G secure channel, wherein The AES encryption algorithm is The plaintext instruction is .

[0059] Further, when the S4 utilizes the all-sensing integration characteristics of the 5G network, the acoustic feedback signals generated by the cloud layer after excitation are captured in real time by the all-sensing antenna array of the 5G base station. The acoustic characteristic spectrum is obtained by filtering and Fourier transform of the signals , wherein The frequency is The time is, and the peak frequency , signal strength and duration are extracted as acoustic evaluation indexes. The optical thickness data of the cloud layer in the intervention process is obtained by the multispectral imager carried by the low-orbit satellite The optical thickness change rate is calculated , wherein The initial optical thickness before intervention is The spectral wavelength is. After the acoustic index and the optical thickness change rate are spatiotemporally aligned, a multi-modal intervention effect evaluation matrix is constructed , wherein the matrix elements are: , wherein And The weight coefficients of the acoustic characteristics and the optical thickness change rate are The acoustic characteristic normalized value of the th time window and the th frequency band is The optical thickness change rate normalized value corresponding to the time window and the wavelength band is The overall strength of the intervention effect is evaluated by the matrix eigenvalue When , the intervention is determined to be effective, wherein

[0060] Furthermore, S5 constructs a cross-regional scheduling computing power pool through inter-satellite laser links between low-Earth orbit (LEO) satellites. Within the LEO satellite constellation, laser communication terminals are mounted to establish communication links with other satellites, achieving a link transmission rate of [missing information]. ,in As the reference rate, Atmospheric attenuation coefficient, To account for the distance between satellites, each satellite will utilize the idle computing power of its computing units. By abstracting virtualization technology into schedulable computing resource blocks, a computing pool is constructed with a total capacity of [missing information]. ,in The number of satellites participating in the network. For the first The computing power utilization coefficient of each satellite is obtained by collecting the computing power requirements of cloud scheduling tasks in various regions through the onboard software-defined network controller. Based on task priority coefficient ,according to Dynamically allocate computing resources to achieve efficient operation of cross-regional computing pool scheduling, among which... For the first The satellite to the first The weighting of each task. This represents the distance between the satellite and the mission area.

[0061] Furthermore, S5 employs a particle swarm optimization algorithm to address the chaotic characteristics of cloud evolution, thereby adjusting the beam illumination time of low-orbit satellites. and the signal transmission power of 5G base stations As optimization variables, construct the particle position vector. ,in For the number of satellites, The number of base stations, with the deviation of the intervention effect. Constructing the fitness function ,in For the first The effect of intervention at any given time For practical results, Initialize particle swarm velocity to determine intervention duration. and location ,pass and Update particles, where For chaotic factors in cloud evolution, For inertial weights, As a learning factor, It is a random number. For the best in particle history, For global optimality, For the chaotic perturbation term, iterate until the condition is met. At that time, the optimal beam illumination time and signal transmission power allocation scheme is output. This is the precision threshold.

[0062] Furthermore, S1 introduces blockchain technology to record the entire scheduling data. It generates encrypted data blocks by timestamping the initial cloud state data, federated learning model parameters, beamforming parameters, intervention commands, and cloud state feedback data collected by low-orbit satellites and 5G networks. Each data block contains the hash value of the previous block. ,pass Construct a chain structure, in which For the current block data, Using timestamps, a distributed network of nodes comprised of satellite constellations and 5G base stations jointly participates in data verification and blockchain accounting to ensure data immutability and traceability. Simultaneously, smart contracts are deployed within the blockchain, with pre-set contract trigger conditions based on the eigenvalues ​​of the multimodal intervention effect evaluation matrix. For multiple consecutive monitoring periods, the value is below the threshold. At that time, the particle swarm optimization algorithm is automatically invoked to regenerate the intervention strategy parameters, and the parameters are sent to the execution terminal through the quantum encryption link. The full-process data after the new strategy is executed is written back into the blockchain to form a new block. Through the cycle of data on-chain - contract monitoring - strategy update - data on-chain again, a self-evolving cloud-oriented scheduling closed loop based on satellite-ground heterogeneous network is formed.

[0063] This implementation details how a high-precision cloud electromagnetic fingerprint map is generated through the collaboration of 3D holographic scanning by a low-Earth orbit (LEO) satellite constellation and a 5G ground-based scatterometer array, addressing the issues of data fragmentation and poor timeliness in existing technologies. At the prediction level, a federated learning framework based on 5G MEC edge nodes, combined with a prediction model featuring transfer learning capabilities, enables accurate prediction of cloud evolution paths in different scenarios, overcoming the limitations of weak adaptability in traditional numerical simulations. At the intervention level, collaborative beamforming technology using LEO satellites and 5G base stations, combined with quantum-encrypted scheduling commands, achieves targeted, precise, and secure intervention on target clouds, overcoming the limitations of limited coverage and low intervention accuracy in existing technologies. At the evaluation and evolution level, a multimodal effect evaluation matrix, a cross-regional computing power pool, and a blockchain closed loop enable real-time quantification of intervention effects and self-evolution of strategies, solving the problems of lagging effect evaluation and lack of dynamic adjustment mechanisms in existing technologies.

[0064] Based on Example 1, this example describes in detail the directional intervention of cumulonimbus clouds based on coordinated beamforming by low-orbit satellites and 5G base stations, specifically as follows:

[0065] like Figure 2(a) As shown, this embodiment is aimed at the scenario of targeted intervention on a cumulonimbus cloud in a certain area, and a distributed transmission array composed of 2 low-orbit satellites and 3 5G base stations is constructed. The low-orbit satellites are equipped with phased array radars and laser communication terminals, and the 5G base stations are configured with integrated sensing and communication antennas. Through the star-ground time synchronization module, the clock deviation is controlled within the preset range. The satellites and base stations are distributed in an equilateral triangle around the target cumulonimbus cloud (three-dimensional coordinates: 30° N, 120° E, 3 km above sea level). The first satellite (S1) is located directly above the target cloud at the zenith, and the second satellite (S2) is at an angle of 60° with S1. The three 5G base stations (G1, G2, G3) are deployed at the three triangular vertices on the ground below the cloud. During system initialization, the distances between each node and the reference point (G1) are obtained through satellite ephemeris and base station GPS positioning data , providing basic parameters for subsequent beamforming weight calculation to ensure the spatial coordination accuracy of the distributed transmission array.

[0066] Based on Figure 2 the spatial coordinates of each node in (b) and the beam pointing angle of the target cumulonimbus cloud , the electromagnetic wave wavelength corresponding to the 26 GHz frequency band, the beamforming weights of each node are calculated through the weight formula : S1's , S2's , G1's , G2's , and G3's . The weight vector is issued through the 5G air interface to control the synchronous transmission of 26 GHz coherent electromagnetic waves by satellites and base stations, forming energy focusing in the target cloud area and ensuring that the electromagnetic waves accurately act on the middle and upper parts of the cumulonimbus cloud.

[0067] As shown in Figure 3 , the scattering echo signals of the cumulonimbus cloud particles are monitored in real time through the satellite hyperspectral imager and the base station sensing antenna, and the resonance peak frequency is extracted through fast Fourier transform. The theoretical resonance frequency of the cumulonimbus cloud particles is = 26.5 GHz, with an allowable deviation of = 0.3 GHz, and the duration threshold of the preset effective intervention window is 5 minutes. In the early monitoring period (0-5 minutes), fluctuates between 26.1-27.0 GHz and does not stabilize in the interval; from the 6th minute, it stabilizes at 26.4-26.6 GHz for 8 minutes, which is determined as an effective intervention window. At this time, the low-orbit satellite generates a scheduling instruction (including beam parameters, action duration, power level, etc.), and generates an encryption key using quantum key distribution technology ,pass ( After being encrypted using the AES encryption algorithm, the data is transmitted to the ground intervention terminal via a 5G secure channel.

[0068] exist Figure 2 Under the collaborative architecture, intervention is executed according to encrypted instructions: maintain initial power for the first 5 minutes. Figure 3 show The signal was stable at 26.5 GHz, indicating good cloud particle resonance. At the 10-minute mark, the satellite observed a dissipation trend in the upper part of the cloud layer. The frequency drops to 26.3 GHz, and the system automatically fine-tunes the transmission phase of G2. Updated to The intervention compensated for path differences caused by changes in cloud height; the intervention ended at 15 minutes, and the echo signal showed... The system remained stable within the target range, and the intervention effect continued. Throughout the process, the satellite-to-ground time synchronization error was controlled within a preset range, ensuring electromagnetic wave coherence; the quantum-encrypted command transmission delay met real-time scheduling requirements. This embodiment verifies the effectiveness of cooperative beamforming technology. Through precise weight calculation, real-time window monitoring, and encrypted command transmission, it achieves directional energy injection into cumulonimbus clouds, providing reliable support for regulating cloud evolution.

[0069] This embodiment achieves targeted and precise intervention in cumulonimbus clouds through coordinated beamforming by low-orbit satellites and 5G base stations. The beam energy is precisely focused, command transmission is secure, and dynamic adjustments are timely, effectively maintaining the resonance state of cloud particles. This verifies the effectiveness of the technology in targeted energy injection and real-time control, laying a reliable foundation for the evolution and control of cumulonimbus clouds.

[0070] Based on Example 1, this example describes in detail the directional intervention of stratocumulus clouds based on coordinated beamforming of low-Earth orbit satellites and 5G base stations, specifically as follows:

[0071] like Figure 4 As shown in (a), this embodiment addresses the directional control scenario of stratocumulus clouds (cloud height 1.5-2km, horizontally uniformly distributed) in a plain area. A distributed transmission array consisting of three low-Earth orbit (LEO) satellites and four 5G base stations is constructed. The LEO satellites (orbit altitude 600km) are equipped with wide-beam phased array radars, and the 5G base stations are configured with omnidirectional sensing antennas (operating frequency band 28GHz). Clock deviations are controlled within a preset range via a satellite-to-ground time synchronization module. The satellites are distributed in a triangular pattern above the stratocumulus clouds, with adjacent satellites at a 90° angle. The four 5G base stations are linearly deployed along the cloud edge, spaced 8km apart, forming a ring-shaped coverage of the stratocumulus clouds. During system initialization, the distance between each node and the reference point (base station G1) is calculated using satellite remote sensing data and base station positioning information. Among them, satellite S1 is at a distance from the reference point = 600.02 km, S2 distance = 600.04 km, S3 distance = 600.03 km, base station G2 distance = 8 km, G3 distance = 16 km, G4 distance = 24 km, providing spatial parameters for beamforming weight calculation, ensuring uniform coverage of large-area stratiform clouds.

[0072] Based on Figure 4 the spatial coordinates of each node in (b), the beam pointing angle of the stratiform cloud (due to the low height of the cloud layer, the beam depression angle is small), the wavelength of the electromagnetic wave (corresponding to the 28 GHz frequency band), according to the weight formula , the beamforming weight of each node is calculated: S1's , S2's , S3's , G1's , G2's , G3's , G4's , the weight vector is issued through the 5G air interface , the satellite transmission power is 25W, the base station transmission power is 8W, the 28GHz coherent electromagnetic wave is synchronously transmitted, and the horizontal uniform distribution of the stratiform cloud is utilized to form a planar energy coverage (power density 0.8W / m²) in the cloud layer area, ensuring that the electromagnetic wave uniformly acts on the lower part of the cloud layer (temperature 2-5°C area).

[0073] As shown in Figure 5 , the satellite multispectral imager and the base station acoustic sensor are used to monitor the scattering echo signal of the stratiform cloud particles in real time, and the resonance peak frequency is extracted by fast Fourier transform. The theoretical resonance frequency of the stratiform cloud particles (diameter 10-20 ) is = 28.2 GHz, the allowable deviation = 0.2 GHz, and the duration threshold of the preset effective intervention window is 10 minutes (because the stratiform cloud evolves slowly); in the initial monitoring period (0-8 minutes), fluctuates between 28.0-28.5 GHz, and is not stable in GHz interval; from the 9th minute, stabilizes at 28.1-28.3 GHz for 12 minutes, which is determined as an effective intervention window. At this time, the low-orbit satellite generates a scheduling instruction (containing beam frequency 28 GHz, action duration 20 minutes, satellite and base station power remaining unchanged), generates an encryption key using quantum key distribution technology, and transmits the key to the base station through the 5G air interface. After encryption by the AES encryption algorithm, the encrypted instructions are transmitted to the ground intervention terminal through the 5G secure channel to ensure that the instruction adaptation layer is consistent with the slow evolution characteristics of the cumulus cloud.

[0074] In the cooperative architecture of the system, the system executes the intervention according to the encrypted instructions: the initial parameters are maintained for the first 10 minutes, Figure 4 The system displays Figure 5 Stable at 28.2 GHz, the resonant state of the cumulus cloud particles is stable; at the 15th minute, the satellite observes that the east of the cloud layer has weak dissipation, Decreased to 28.1 GHz, the system automatically adjusts the transmission phase of S2 and G3 to update , Updated to , Updated to , to compensate for the energy distribution deviation caused by the horizontal drift of the cloud layer; at the 20th minute, the intervention ends, and the echo signal shows Still stable in the target interval, the overall structure of the cloud layer remains intact. During the entire process, the satellite-ground time synchronization error is controlled within the preset range, ensuring the coherence of the electromagnetic wave; the quantum encryption instruction transmission delay is <150ms, meeting the real-time requirements of the slow intervention of the cumulus cloud. This embodiment verifies the applicability of the cooperative beam forming technology to the cumulus cloud, and through the parameter design and dynamic adjustment of the cloud layer characteristics, the directional and stable intervention of the cumulus cloud is realized, further proving that the application can cover the regulation and control requirements of different types of cloud layers.

[0075] This embodiment is based on the uniform distribution characteristics of the cumulus cloud, and the surface energy coverage is realized through cooperative beam forming. The resonant frequency is stable in the effective interval, continuously meeting the intervention window requirements, and dynamically fine-tuned to compensate for the cloud drift, verifying the adaptability of the technology to slow-evolving clouds, proving that it can stably regulate and control the cumulus cloud, and embodying the universality.

[0076] Based on embodiment 1, this embodiment describes the cross-regional cloud layer scheduling computing power allocation based on the low-orbit satellite inter-satellite laser link and particle swarm optimization, specifically:

[0077] For example Figure 6 ​As shown in (a), for a multi-regional mixed scheduling scenario of stratocumulus and cumulonimbus clouds, an inter-satellite laser link network consisting of four low-Earth orbit satellites (S1-S4, orbital altitude 550km) is constructed to form a cross-regional computing power pool covering 110°-120°E and 20°-30°N. The satellites are arranged in a square orbit with an adjacent satellite spacing of 1000km. Each satellite carries a laser communication terminal (transmission wavelength 1550nm) and a 50TOPS computing unit, realizing the virtualization and integration of computing resources through inter-satellite laser links. Among them, S1 covers the cumulonimbus cloud region in South China, S2 covers the stratocumulus cloud region in Central China, and S3 and S4 serve as redundant nodes to undertake load balancing tasks; the link transmission rate is calculated according to... Calculation, reference rate =10Gbps, atmospheric attenuation coefficient The actual transmission rate between S1 and S2 is Each satellite will utilize its idle computing power (S1=20TOPS, S2=15TOPS, S3=18TOPS, S4=22TOPS) and computing power utilization coefficient (0.8-0.9) Abstracted into resource blocks using virtualization technology, the total computing power pool capacity is calculated according to... Calculated It provides computing power support for cross-regional task allocation.

[0078] Figure 6 (b) marks three cloud scheduling tasks with different priorities: T1 (rain enhancement from cumulonimbus clouds in South China, priority) =0.9, computing power requirement =12TOPS), T2 (dissipation of stratocumulus clouds in Central China, priority) =0.7, computing power requirement =8TOPS), T3 (coastal mixed cloud monitoring, priority) =0.5, computing power requirement =5TOPS), according to the weighting formula Calculate the weight assigned to each satellite for the mission: distance from S1 to T1 Distance from S2 to T2 Distance from S3 to T3 The weight calculation results are transformed into actual computing power allocation through the onboard software-defined network controller: T1 is jointly carried by S1 (8TOPS) and S4 (4TOPS), T2 is carried by S2 (6TOPS) and S3 (2TOPS), and T3 is carried solely by the remaining computing power of S3 (10TOPS), realizing the scheduling logic of "near-domain priority and load balancing". Figure 6 The direction of computing power flow indicated by the thickness of the arrows in (a) and (b) is consistent.

[0079] The particle swarm optimization process is irradiated by low-orbit satellite beam illumination time (S1-S4 - , range 5-15 minutes) and 5G base station transmission power (10 base stations - , range 10-30W) as optimization variables, and a particle position vector is constructed ; as shown in Figure 7 , a fitness function is constructed according to , wherein the intervention effect deviation ( the target effect , the actual effect ). The algorithm parameters are set as follows: inertia weight =0.7, learning factor = =1.5, chaos factor =0.3, and precision threshold =0.001. 50 particles are initialized, and the position is updated by and , and the blue curve in Figure 6 shows that the fitness function value converges to 0.92 at the 35th iteration, satisfying , and the optimal parameter combination is output.

[0080] Figure 7 The parameter optimization results on the right show that the S1 beam irradiation time for cumulonimbus T1 is optimized to =12 minutes, and the 5G base station - =25W (because the cumulonimbus particle concentration is high, strong power focusing is required); the S2 irradiation time for stratocumulus T2 is optimized to =8 minutes, and the base station - =15W (the stratocumulus is thin, and the power is moderately reduced); the S3 irradiation time for T3 is =10 minutes, and the base station - =20W (balance monitoring and intervention needs), and the chaos disturbance term (taking values 0.1-0.5) avoids the particles from falling into local optima, making the irradiation time and power distribution present a "chaotic order" rule—the cumulonimbus intervention time is concentrated in 10-15 minutes, and the stratocumulus is concentrated in 5-10 minutes, matching the difference in evolution rate of the two types of clouds. Figure 6 The convergence trend of the global optimal solution (gbest) and the particle historical optimal solution (pbest) marked by the red dashed line in verifies the adaptability of the algorithm to complex cloud systems.

[0081] The embodiment verifies the effect through three groups of comparison experiments: (1) when only the computing power pool is enabled without optimizing the parameters, the T1 task completion delay is 18 minutes, and the effect deviation E=5.2; (2) when only the parameters are optimized without enabling the computing power pool, the T2 task is interrupted due to insufficient computing power, and after restarting, the delay increases by 40%; (3) when the computing power pool and the particle swarm optimization are cooperatively enabled, the average completion delay of T1-T3 is reduced to 9 minutes, the effect deviation E=1.8, which meets the effective intervention standard. The computing power pool of the inter-satellite laser link solves the problem of "regional insufficient computing power", and the particle swarm optimization algorithm solves the problem of "poor parameter adaptability", and the combination of the two makes the efficiency of cross-regional cloud layer scheduling improve by more than 60%.

[0082] The embodiment builds a cross-regional computing power pool through an inter-satellite laser link, and realizes dynamic allocation of computing power by combining a particle swarm optimization algorithm, the computing power scheduling delay is reduced by 60%, the task completion efficiency is improved by 40%, the fitness function converges to 0.92, and the accuracy threshold is met. Different types of cloud layer tasks are supported by accurate computing power, verifying the synergistic effect of inter-satellite cooperation and intelligent optimization, solving the problem of uneven regional computing power, and providing efficient and stable technical support for cross-domain scheduling of complex cloud layer systems.

[0083] The above is only a preferred embodiment of the present application, and does not limit the protection scope of the present application. For those skilled in the art, the present application can have various changes and variations; any change, modification, replacement, integration and parameter change of these embodiments within the spirit and principles of the present application, which can realize the same function without departing from the principles and spirit of the present application, falls within the protection scope of the present application.

Claims

1. A cloud layer directional scheduling method based on low earth orbit satellites and 5G networks, characterized in that, Comprise: S1, through the low-orbit satellite constellation to build dynamic cloud topology aware network, through the satellite phased array radar to carry on the three-dimensional holographic scanning to the cloud layer, extracts the electromagnetic scattering characteristic spectrum of cloud particle, utilizes the ultra wide band characteristic of 5G air interface, the cloud microwave response data of ground scatterometer array collection and satellite data are carried out cross-domain fusion, generate cloud electromagnetic fingerprint atlas; S2, based on 5G MEC edge node deployment federal learning framework, each node utilizes the historical cloud scheduling case stored locally to train submodel, realizes model parameter encryption aggregation through satellite link, constructs the cloud behavior prediction model with the ability of transfer learning, outputs the nonlinear evolution path of cloud under different intervention intensity; S3, adopt the cooperative beam forming technology of low-orbit satellite and 5G base station, to the target cloud layer emits specific frequency band coherent electromagnetic wave, through monitoring the cloud particle resonance frequency change judges the intervention window, generates the scheduling instruction based on quantum key encryption; S4, utilize the 5G network's integrated feature, real-time capture the acoustic feedback signal of cloud after excitation, combine the cloud optical thickness change observed by satellite, build multi-modal intervention effect evaluation matrix; S5, through the low-orbit satellite inter-satellite laser link to build cross-regional scheduling computing power pool, based on the chaos characteristic of cloud evolution adopts particle swarm optimization algorithm, dynamically allocates the beam irradiation time of different satellites and the signal transmission power of 5G base station; S6, introduce the block chain technology to record the whole process scheduling data, utilize the smart contract to automatically trigger the iteration update of cloud intervention strategy, form the self-evolution cloud directional scheduling closed loop based on satellite-ground heterogeneous network. 2.The cloud tier-directed scheduling method based on low earth orbit satellites and 5G networks according to claim 1, characterized in that, When the dynamic cloud layer topology perception network is constructed by the low-orbit satellite constellation in the S1, multiple low-orbit satellites form a distributed observation array according to preset orbits, phased array radars carried by the satellites perform fan beam scanning on a target airspace in a time division multiplexing manner, the azimuth angle and the elevation angle of the radar beam are adjusted to realize three-dimensional airspace coverage , and after pulse compression processing is performed on echo signals obtained through scanning, scattering intensity of cloud layer particles at different frequency bands is extracted , is a radar transmission frequency, and a propagation attenuation coefficient of a radar wave in a cloud layer is combined to obtain an electromagnetic scattering characteristic spectrum of the cloud layer particles through calculation of a scattering cross section , wherein is an equivalent radius of the particles, is a distance from the satellite to the cloud layer, is a transmission power, , are respectively a transmission antenna gain and a receiving antenna gain, is a radar wavelength.​​ 3.The cloud tier-directed scheduling method based on low earth orbit satellites and 5G networks according to claim 1 or 2, characterized in that, The S1 utilizes the ultra-wideband characteristics of the 5G air interface to realize multi-frequency band synchronous observation of the cloud layer through the ultra-wideband carrier aggregation technology of the 5G NR on the ground scatterometer array in a preset frequency band, and collect microwave response data of the cloud layer acquired by scatterometers at different positions wherein is the scatterometer number, is the observation time, and is the electromagnetic scattering characteristic spectrum of the satellite, is spatiotemporally aligned with the ground data, a fusion weight matrix is constructed, a covariance correction term is used to dynamically adjust the weight, wherein , are weight coefficients of the satellite data and the ground data, respectively, is a frequency point index, , are mean values of the satellite data and the ground data, respectively, is a variance, and a cloud electromagnetic fingerprint spectrum is generated, and the formula is: satellite and ground data cross-domain fusion is realized. 4.The cloud tier-directed scheduling method based on low earth orbit satellites and 5G networks of claim 1, wherein, In S2, when deploying the federated learning framework based on 5G MEC edge nodes, federated learning clients are deployed in each 5G MEC edge node within the target area. Each client calls the locally stored historical cloud scheduling case dataset. ,in For the first The node of the first Each case input feature includes the initial cloud state and intervention parameters. To correspond with the output results, a local sub-model is trained using the stochastic gradient descent algorithm. Calculate the gradient ,in For the first The loss function for each node. For single sample loss, For the first The sample size of each node is [number], and after training, each node encrypts the sub-model parameters using a symmetric encryption method based on the SM4 algorithm via a 5G network. The encryption parameters are [details]. ,in The session key, pre-distributed via satellite link, is sent to a low-Earth orbit satellite acting as an aggregation server. The satellite receives the encryption parameters from all nodes and decrypts them to obtain the desired key. The aggregation model is calculated using the following formula: ,in The total number of edge nodes. For the first The weight coefficients of the sub-model parameters of each edge node during aggregation, and the weight coefficients of the sub-model parameters during aggregation, and the weight coefficients of the sub-model parameters of each edge node ... After encryption, the data is broadcast to each 5G MEC edge node via satellite link to complete the sub-model update. 5.The cloud tier-directed scheduling method based on low earth orbit satellites and 5G networks according to claim 1 or 4, characterized in that, The S2 is configured to have a cloud layer behavior prediction model with migration learning ability The basic model is used as a feature mapping function The source domain data features And the target domain data features Map to a shared feature space Calculate the domain difference loss The formula is: Wherein The domain discriminator, The domain label, , The source domain and the target domain sample size are respectively weighted and summed with the prediction loss To get the total loss Wherein The model prediction value, The measured value, The weight coefficient is optimized by minimizing The model parameters, the model input is different intervention intensity Wherein The electromagnetic wave power corresponding to the first The action time, the output is the nonlinear evolution path of the cloud layer at future Time Wherein The position coordinates, The thickness, The particle concentration. 6.The cloud tier-directed scheduling method based on low earth orbit satellites and 5G networks according to claim 1, wherein, In S3, when using the collaborative beamforming technology of low-Earth orbit satellites and 5G base stations, the low-Earth orbit satellites and the 5G base stations covering the target cloud area form a distributed transmission array. The transmission clock deviation between the satellites and base stations is controlled within a preset time range through a satellite-to-ground time synchronization module. The beamforming weight vector of each transmission node is calculated based on the three-dimensional coordinates of the target cloud layer. ,in The total number of collaborating nodes is given by the formula: , For the first The distance between each node and the reference point Beam pointing angle, (Wavelength of electromagnetic waves), controlling each node to synchronously emit coherent electromagnetic waves of different frequency bands and form energy focusing in the target cloud region, real-time monitoring of the scattered echo signals of electromagnetic waves of different frequencies by cloud particles, and extracting the resonant peak frequency in the echo spectrum through fast Fourier transform. ,when Stable within the preset frequency range ,middle The theoretical resonant frequency, To determine an effective intervention window, a deviation value must be allowed, and the duration of the deviation must exceed a threshold. A scheduling command containing beam parameters, duration of action, and power level is generated by the low-Earth orbit satellite, and an encryption key is generated using quantum key distribution technology. ,pass After the command is encrypted, it is transmitted to the cloud-based intervention and execution terminal via a 5G secure channel. It uses the AES encryption algorithm. This is a plaintext instruction. 7.The cloud tier-directed scheduling method based on low earth orbit satellites and 5G networks according to claim 1, wherein, The S4 utilizes the 5G network's sensing integration feature, and the acoustic feedback signal generated by the cloud layer after excitation is captured in real time by the sensing antenna array of the 5G base station, and the signal is filtered and Fourier transformed to obtain the acoustic feature spectrum wherein is the frequency, is the time, and the peak frequency , signal intensity and duration in the feature spectrum are extracted The optical thickness data of the cloud layer during the intervention process is obtained by a multispectral imager carried by a low-orbit satellite as an acoustic evaluation index , and the optical thickness change rate is calculated wherein is the initial optical thickness before intervention, is the wavelength, and after the acoustic index and the optical thickness change rate are spatiotemporally aligned, a multi-modal intervention effect evaluation matrix is constructed wherein the matrix elements are: wherein and are the weight coefficients of the acoustic features and the optical thickness change rate, respectively, is the acoustic feature normalized value of the th time window and the th frequency band, is the optical thickness change rate normalized value corresponding to the time window and the wavelength band, and the overall strength of the intervention effect is evaluated by the matrix eigenvalue , and when is the preset threshold value. 8.The cloud tier-directed scheduling method based on low earth orbit satellites and 5G networks of claim 1, wherein, In S5, a cross-regional scheduling computing power pool is constructed through inter-satellite laser links between low-Earth orbit (LEO) satellites. Within the LEO satellite constellation, laser communication terminals are mounted to establish communication links with other satellites, with a link transmission rate of [missing information]. ,in As the reference rate, Atmospheric attenuation coefficient, To account for the distance between satellites, each satellite will utilize the idle computing power of its computing units. By abstracting virtualization technology into schedulable computing resource blocks, a computing pool is constructed with a total capacity of [missing information]. ,in The number of satellites participating in the network. For the first The computing power utilization coefficient of each satellite is obtained by collecting the computing power requirements of cloud scheduling tasks in various regions through the onboard software-defined network controller. Based on task priority coefficient ,according to Dynamically allocate computing resources to achieve efficient operation of cross-regional computing pool scheduling, among which... For the first The satellite to the first The weighting of each task. This represents the distance between the satellite and the mission area. 9.The cloud tier-directed scheduling method based on low earth orbit satellites and 5G networks according to claim 1 or 8, characterized in that, The chaos characteristic based on cloud evolution in S5 adopts a particle swarm optimization algorithm, and the beam irradiation time of a low-orbit satellite is calculated And the signal transmission power of a 5G base station As an optimization variable, a particle position vector is constructed , wherein is the number of satellites, is the number of base stations, and the intervention effect deviation is constructed , wherein is the target intervention effect at the moment, is the actual effect, is the intervention duration, the particle swarm velocity and position are initialized and are updated, wherein is a cloud evolution chaos factor, is an inertia weight, is a learning factor, is a random number, is a particle historical optimum, is a global optimum, is a chaotic disturbance term, and iteration is performed until is satisfied, and the optimal beam irradiation time and signal transmission power distribution scheme are output, is a precision threshold. 10.The cloud tier-directed scheduling method based on low earth orbit satellites and 5G networks according to claim 1, wherein, S1 introduces blockchain technology to record the entire scheduling data. The initial cloud state data, federated learning model parameters, beamforming parameters, intervention commands, and cloud state feedback data collected by low-orbit satellites and 5G networks are used to generate encrypted data blocks according to timestamps. Each data block contains the hash value of the previous block. ,pass Construct a chain structure, in which For the current block data, Using timestamps, a distributed network of nodes comprised of satellite constellations and 5G base stations jointly participates in data verification and blockchain accounting to ensure data immutability and traceability. Simultaneously, smart contracts are deployed within the blockchain, with pre-set contract trigger conditions based on the eigenvalues ​​of the multimodal intervention effect evaluation matrix. For multiple consecutive monitoring periods, the value is below the threshold. At that time, the particle swarm optimization algorithm is automatically invoked to regenerate the intervention strategy parameters, and the parameters are sent to the execution terminal through the quantum encryption link. The full-process data after the new strategy is executed is written back into the blockchain to form a new block. Through the cycle of data on-chain - contract monitoring - strategy update - data on-chain again, a self-evolving cloud-oriented scheduling closed loop based on satellite-ground heterogeneous network is formed.

Images